[OSM-talk] Fixme: A proposal

Jochen Topf jochen at remote.org
Mon Oct 3 07:48:51 BST 2011


On Sun, Oct 02, 2011 at 05:29:04PM -0700, John Harvey wrote:
> I have an experimental renderer and from time to time I find "bugs"
> with planet.osm that I would like to see fixed.  From what I can
> tell, I'm not alone - GeoFabrik
> (http://wiki.openstreetmap.org/wiki/OSM_Inspector) Skobbler
> (http://wiki.openstreetmap.org/wiki/Mapdust) and
> http://wiki.openstreetmap.org/wiki/Keep_Right are years ahead of me
> in reporting bugs, but I suspect I detect some novels bugs that
> aren't frequently reported.
> 
> I like to contribute content in my area, and I am willing to take
> time to fix bugs in my area.  Trying to find local issues isn't
> obvious - there are places to find issues, but they certainly aren't
> obvious from Potlatch.
> 
> I would like to propose three changes:
> 
>  * We currently have roughly 700,000 nodes/ways/relations marked with
>    FixMe.  There isn't a lot of structure here, and I suspect a lot of
>    these issues aren't moving.  I propose we add a layer of
> structure.    Would could categorize bugs into categories.  For
> instance:
>      o FixMe:Legal - Content with potentially legal issues
>      o FixMe:Topology - Content that violates the semantics of how it
>        is tagged.  (Multigons with inners outside of all outers for
>        instance).
>      o FixMe:Redundant - overlapping geometry, or tags which are redundant.
>      o FixMe: Nice to have and so on
> 
>    I think the goal should be for the 700K existing items to eventually
>    be categorized and the "error" type issues get fixed.

Problems with the data that are automatically detected should never ever be
tagged on the data itself. This is useless, and worse, its counterproductive.
Automatically detected problems should be flagged in specialized tools, you
mentioned some of them. When they are fixed, they go away. If they are tagged
with special fixme tags you now have four cases (ok and not marked, ok and
marked as bug, not ok and marked as bug and not ok and not marked as bug)
instead of two (ok or not). Unfortunately this is a common practice. There
are hundreds of thousands of those cases that obscure the more important
problems.

Only manually detected problems should be tagged with fixme. And it should be
as easy as possible to mark those. Extra categorization would make it more
difficult. Why exactly do you want the extra categorization? What help would it
be in your practical day to day work?

I can see one useful differentiation: Some problems are fixable only with local
knowledge (say a missing street name), some are fixable from afar (most
topological problems). It might be helpful to not see problems needing local
knowledge in areas where I don't have local knowledge.

>  * People in their Profile Description identify regions for which they
>    will accept notifications of newly detected bugs, even if they
>    didn't author the content.  For instance:
>      o FixMeOwn: -123.321 -122.959 49.204 49.328
> 
>    As a contributor, if I was told about new issues in my area of
>    concern, I would take a look at them (no promises I can fix them).

That idea or variants have been floated several times. Yes, it would be nice
to have something like this. Just needs somebody to implement it. :-)

>  * I believe that the tiles rendered on the www.openstreetmap.org page
>    should be marked up to show errors (not all fixme's, but a known set
>    of them).  The tiles on www.openstreetmap.org are there for mappers
>    to fix data, not to just look pretty.  A red outline on ways/nodes
>    that have FixMe:Legal, FixMe:Topology and FixMe:Vandalism would draw
>    mappers attention to things that can improve.

We have been through many rounds of this discussion: Is the map on the main
OSM page more for *users* of the data or more for *mappers*. There are many
pros and cons.

You can switch to the NoName layer which does a bit what you are proposing.
But it always has been the "odd one out", just one layer showing one particular
problem. Maybe its better to stick with the specialized tools that can do a
better job at helping the mapper with fixing bugs.

> I'm just floating an idea.  I can do very little to forward these
> ideas, but I can tell you I would contribute more if there was a way
> for me to enter bugs I find world wide and find bugs in the areas I
> know.

Jochen
-- 
Jochen Topf  jochen at remote.org  http://www.remote.org/jochen/  +49-721-388298




More information about the talk mailing list